Struct AudioSampling 

pub struct AudioSampling {
    pub temperature: f32,
    pub top_k: usize,
    pub repetition_penalty: f32,
    pub repetition_window: usize,
    pub fixed_order: bool,
}

Audio sampling: Liquid’s recommended audio_temperature 1.0, audio_top_k 128, repetition penalty 1.1 over the last 4096 frames.

Fields

temperature: f32

top_k: usize

repetition_penalty: f32

repetition_window: usize

fixed_order: bool

Reveal the codebooks in order 0..11 instead of a random order (the reference harness’s fixed order).

Implementations

impl AudioSampling

pub fn from_model(model: &CmfModel) -> Self

The release’s own audio_* settings (its generation config, kept by the converter in provenance.lfm3_audio.generation): LFM3-Omni samples at 0.4 with a 1.2 penalty over 64 frames and no top-k.

Trait Implementations

impl Clone for AudioSampling

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AudioSampling

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AudioSampling

fn default() -> Self

Returns the “default value” for a type.